Published in "Oecologia 191(1): 97–112, 2019"

which should be cited to refer to this work.

Kin-dependent dispersal influences relatedness and genetic

structuring in a lek system
Hugo Cayuela1  · Laurent Boualit2 · Martin Laporte1 · Jérôme G. Prunier3 · Françoise Preiss4 · Alain Laurent4 ·
Francesco Foletti2 · Jean Clobert3 · Gwenaël Jacob2

Abstract
Kin selection and dispersal play a critical role in the evolution of cooperative breeding systems. Limited dispersal increases
relatedness in spatially structured populations (population viscosity), with the result that neighbours tend to be genealogi-
cal relatives. Yet the increase in neighbours’ fitness-related performance through altruistic interaction may also result in
http://doc.rero.ch

habitat saturation and thus exacerbate local competition between kin. Our goal was to detect the footprint of kin selection
and competition by examining the spatial structure of relatedness and by comparing non-effective and effective dispersal in
a population of a lekking bird, Tetrao urogallus. For this purpose, we analysed capture–recapture and genetic data collected
over a 6-year period on a spatially structured population of T. urogallus in France. Our findings revealed a strong spatial
structure of relatedness in males. They also indicated that the population viscosity could allow male cooperation through
two non-exclusive mechanisms. First, at their first lek attendance, males aggregate in a lek composed of relatives. Second,
the distance corresponding to non-effective dispersal dramatically outweighed effective dispersal distance, which suggests
that dispersers incur high post-settlement costs. These two mechanisms result in strong population genetic structuring in
males. In females, our findings revealed a lower level of spatial structure of relatedness and genetic structure in respect to
males. Additionally, non-effective dispersal and effective dispersal distances in females were highly similar, which suggests
limited post-settlement costs. These results indicate that kin-dependent dispersal decisions and costs have a genetic footprint
in wild populations and are factors that may be involved in the evolution of cooperative courtship.

Keywords Cooperation · Sociality · Dispersal · Relatedness · Genetic structuring · Bird · Tetrao urogallus

Hugo Cayuela and Laurent Boualit contributed equally.
* Hugo Cayuela
hugo.cayuela51@gmail.com
1
Département de Biologie, Institut de Biologie Intégrative
et des Systèmes (IBIS), Université Laval, Pavillon Charles-
Eugène-Marchand, Québec, QC G1V 0A6, Canada
2
Department of Biology, University of Fribourg, Fribourg,
Switzerland
3
Theoretical and Experimental Ecology Station (UMR 5371),
National Centre for Scientific Research (CNRS), Paul
Sabatier University (UPS), Moulis, France
4
Groupe Tétras Vosges, Maison du Parc, 1, cour de l’Abbaye,
68140 Munster, France
Introduction
Kin selection plays a critical role in the evolution of coop-
erative breeding systems (Clutton-Brock 2002; Griffin
and West 2003; Bourke 2011). In these systems, breeding
individuals are assisted by cooperative individuals (also
called helpers) that apparently sacrifice part or all of their
own reproductive potential. An explanation for this appar-
ent paradox is put forward by Hamilton’s theory (1964):
because relatives share genes in common, if an individual
improves the reproductive success of its relatives, this may
increase its overall genetic contribution to future genera-
tions (i.e. inclusive fitness). In turn, genes that promote
cooperation may then increase in frequency in the popula-
tion. Accordingly, relatedness seems an important factor
for the evolution of cooperative behaviour, implying that
individuals should tend to interact within family groups or
with other genetically similar individuals (Hamilton 1964;

1
 El Mouden and Gardner 2008; Platt and Bever 2009; but
also see Clutton-Brock 2009). Two alternative mechanisms
have been proposed regarding how an individual ensures
that cooperative behaviour is directed primarily towards
social partners with high relatedness. The first is the abil-
ity to recognize and act preferentially towards genealogi-
cal kin (kin discrimination), which requires odour-based
recognition systems or sophisticated cognitive faculties
(Hepper 2005; Komdeur et al. 2008). The second is lim-
ited dispersal, which drastically increases relatedness in
spatially structured populations (population viscosity)
(Hamilton 1964; El Mouden and Gardner 2008; Platt and
Bever 2009), with the result that neighbours tend to be
genealogical relatives. Yet the increase in neighbours’
performance through altruistic interaction may also result
in habitat saturation and thus exacerbate local competi-
tion between kin (Queller 1992; Platt and Bever 2009).
http://doc.rero.ch
In general, the evolution of sociality is based on a bal-
ance between kin cooperation and kin competition, which
depends on a species’ life history traits, the spatial scale
over which cooperation and competition occur, and the
underlying habitat structure (Queller 1992; Le Galliard
et al. 2003, 2004).
Dispersal designates the movement of an individual from
its birth site to its reproduction patch, as well as between
successive reproduction sites (Clobert et al. 2009; Matthy-
sen 2012). It is usually considered a three-stage process that
includes departure (or emigration), transience (or transfer
within the landscape matrix) and settlement (or immigra-
tion) (Ronce 2007; Baguette et al. 2013). A recurrent find-
ing of theoretical models is that dispersal evolution is based
on a trade-off between related costs (i.e. energetic costs,
time costs, risk costs and opportunity costs) and benefits
at each stage of the process (Bonte et al. 2012). In coopera-
tive breeding systems, relatively high dispersal rates have
been reported by some field studies (e.g. Greenwood and
Harvey 1982; Clarke et al. 1997), a phenomenon that should
decrease the relatedness level within a patch, reduce the vis-
cosity of the whole population and constrain the evolution of
cooperative behaviour. Two explanations may be possible for
this apparent paradox: first, it is conceivable that population
viscosity is maintained if dispersers pay acute dispersal costs
during transience and after settling into a new patch (Forero
et al. 2002; Hansson et al. 2004; Nystrand 2007; Dickinson
et al. 2009), limiting their contribution to local reproduction.
Second, cooperative behaviour may evolve if dispersers pref-
erentially join patches composed of close relatives (Sinervo
and Clobert 2003; Piertney et al. 2008; Rosher et al. 2017).
To date, little is known about the consequences of dispersal
costs and kin-dependent dispersal on gene flow and popula-
tion viscosity, and therefore about their contribution to the
evolution of cooperative behaviour, especially in the context
of lek systems.
2
Lek mating systems are an interesting case study for
investigating this issue. In lek-breeding species, males
congregate on display grounds during the breeding period,
allowing females to compare potential partners before solic-
iting mating. A female’s fitness is mainly limited by her own
reproductive investment. Females are thus under selection
pressure to reliably recognise high-quality partners, which
in turn imposes a sexual selection pressure on males to
advertise their quality through costly ornamental traits and
courtship (Møller and Alatalo 1999; Tregenza and Wedell
2000). This leads to intense intrasexual competition and
strongly skewed reproductive success in males (Kokko and
Lindstrom 1996; Kokko et al. 1999; Höglund and Alatalo
2014; Verkuil et al. 2014). In several lek-breeding birds,
theoretical studies have demonstrated that low-quality males
might benefit from indirect benefits of kin selection through
increasing lek size (Kokko and Lindstrom 1996). Empirical
studies have shown that females preferentially mate in large
leks (Alatalo et al. 1992). By joining a lek occupied by rela-
tives, males with low reproduction prospects may enhance
the overall attractiveness of the lek and therefore increase
the fitness of related dominant males (Höglund 2003). The
males unsuccessful in obtaining copulations may themselves
benefit from the resulting inclusive fitness (Krakauer 2005;
DuVal 2007). Theoretically, limited dispersal between leks
should ensure a sufficient level of population viscosity to
favour the evolution of cooperative courtship behaviour.
However, field studies have revealed that dispersal can some-
times be high (Höglund and Alatalo 2014), with no clear
indication that dispersal is counter-selected. In this context,
a higher level of population viscosity could be maintained:
(1) if dispersers preferentially join leks composed of close
relatives, and/or (2) if dispersers incur sufficient reproduc-
tive costs during transience and/or after settling into a new
lek consisting of non-relatives.
An elegant way to evaluate the reproductive costs of
dispersal in the wild is by comparing ‘non-effective’ and
‘effective’ dispersal. Non-effective dispersal describes dis-
persal events that may or may not be followed by successful
reproduction (Broquet and Petit 2009; Cayuela et al. 2018a,
b). This kind of dispersal is usually quantified by means
of direct observations in the field: for instance, using cap-
ture–recapture methods (Lebreton et al. 2009; Lowe and
Allendorf 2010). In contrast, effective dispersal describes
dispersal events followed by successful reproduction and
resulting in gene flow. Effective dispersal can be quantified
using molecular approaches (Prugnolle and De Meeûs 2002;
Broquet and Petit 2009). When no dispersal cost occurs or
this cost is compensated (Cotto et al. 2014), one would
expect similar non-effective and effective rates of dispersal,
over a similar range of distance. In contrast, when disper-
sal entails immediate or delayed costs resulting in a loss
of reproductive success, one would expect non-effective
 dispersal to exceed effective dispersal (Prugnolle and De
Meeûs 2002; Broquet and Petit 2009).
In this study, we aimed at detecting the footprint of kin
selection and competition by examining the spatial structure
of relatedness and by comparing non-effective and effective
dispersal distances in a lekking bird, the Western capercail-
lie (Tetrao urogallus). In tetraonids, subordinate males (i.e.,
helpers) aggregate in leks in which few dominant relatives
acquire most of the mates (Höglund and Alatalo 2014), a
breeding system that shows some similarities with the coop-
erative breeding system described in wild turkeys (Krakauer
2005). We used capture–recapture and genetic data collected
over a 6-year period (corresponding to two generations of
capercaillie in our study system) in a spatially structured
population of T. urogallus located in the Vosges Mountains
of eastern France. Previous studies have highlighted a relat-
edness structure in males of this species at a small spatial
http://doc.rero.ch
scale, possibly driven by kin selection (Regnaut et al. 2006;
Segelbacher et al. 2007). We hypothesized that male popula-
tion viscosity is potentially maintained by two non-exclusive
mechanisms: philopatry or high post-settlement costs, both
of which are likely to enhance the genetic structure of a
population. The first mechanism implies context-dependent
dispersal, with males preferentially attending leks com-
posed of relatives. At their first attendance to a lek, young
males join one consisting of relatives (e.g., father, full- or
half-sibling) and then exhibit philopatric behaviour due to
kin-selected benefits and other social and environmental
factors (e.g., position in the local dominance hierarchy or
intrasexual competition for a territory). Alternatively, they
may leave the lek (or attend another one) to avoid kin com-
petition, which should contribute to a decrease in population
viscosity. The second mechanism allowing male population
viscosity to be maintained could be high post-settlement
costs due to intense intrasexual competition and the loss of
kin-selected benefits. Given these hypotheses, we expected
that non-effective dispersal would exceed effective dispersal
(i.e. gene flow). We also examined female non-effective and
effective dispersal, relatedness spatial structure and genetic
structuring. We expected a lower relatedness structure in
females than in males, likely due to inbreeding avoidance
(Pusey 1987; Szulkin and Sheldon 2008) and the absence
of kin-selected benefits (Höglund and Alatalo 2014). Young
females may thus avoid leks composed of relatives, and
older females may emigrate from leks composed of relatives.
As female–female competition for mate acquisition is mar-
ginal in lek systems (Höglund and Alatalo 2014), we also
expected limited post-settlement costs for females arriving
to mate in a new lek; accordingly, we hypothesized that non-
effective and effective dispersal for females would be rela-
tively similar. In both males and females, we inferred effec-
tive dispersal distances uncorrected (using mantel tests) and
corrected (multiple regressions) for landscape connectivity.
3
Materials and methods
Study area, sampling design and identification
of individuals
The study was conducted on a spatially structured popu-
lation of T. urogallus located in the Vosges Mountains
of eastern France (Fig. 1). Between 2010 and 2015, the
collection of non-invasive samples (95% faeces and 5%
feathers) allowed us to identify 109 individuals (61 males
and 48 females) distributed among eleven leks (1–18 indi-
viduals identified par year and per lek).
We do not have behavioral data showing that males pre-
sent at a lek are territorial and attempt to reproduce. How-
ever, the lekking behaviour is energetically costly and the
lek zone does not seem to be used for other activities than
reproduction (Höglund and Alatalo 2014). It is therefore
very likely that the presence of male at a lek indicates that
it actively participates to the reproduction in this lek. We
therefore assumed that a male attended a given lek when
its faeces or feathers were found at least once in this lek.
All known active leks within the study area were
exhaustively sampled (Fig. 1). A detailed description of
the study area, the sampling method and the genotyping
approach is provided in Appendix A in ESM. The Euclid-
ean distance between leks ranged from 2.6 to 42 km, with
a median of 18 km. The population was surveyed using
the capture–recapture (CR) method over a 6-year period
(2010–2015).
Estimating non-effective dispersal from capture–
recapture data
We quantified dispersal rates and distances for both sexes.
Non-effective dispersal rate per generation and dispersal
distance were estimated separately to facilitate model
implementation. The models used in these analyses are
described below. Prior to build capture-recapture mod-
els, we carried out Goodness-of-Fit tests (GOF). As no
GOF test exists for multievent capture-recapture models,
we used GOF tests designed for Cormack–Jolly–Seber
models. We ran separate analyses for the two sexes. We
examined the presence of transience and trap-dependence
using the GOF tests implemented in the program UCARE
(Choquet et  al. 2009a). These analyses revealed the
absence of transience in both females (overall test, df = 4,
𝜒 2 = 2.44, p = 0.65) and males (df = 4, 𝜒 2 = 1.22, p = 0.87).
Similarly, we did not detect trap-dependence in females
(df = 3, 𝜒 2 = 0.54, p = 0.91) and males (df = 3, 𝜒 2 = 0.52,
p = 0.92). Therefore, we did not consider transience and
trap-dependence in our capture-recapture models.
 Fig. 1 Map of the study area
in the Vosges Mountains of
eastern France. The 11 leks
surveyed using capture–recap-
ture and genetic methods are
indicated according to size in
terms of number of individu-
als. Each lek was identified by
a three-letter code. The total
number of individuals captured
at least once over the 5-year
study period was: GdF = 57,
NOI = 26, CrH = 16, StA = 15,
LLC = 31, GEH = 10, TdR = 5,
BAG = 6, TAN = 13, FOS = 11,
VEN = 46
http://doc.rero.ch
Quantifying the non-effective dispersal rate per generation
We modelled dispersal using multievent CR models (Pradel
2005); more specifically a model proposed by Lagrange
et al. (2014) that allows the estimation of survival (ϕ) and
dispersal (ψ) between numerous sites (or leks in our study
case). By omitting site identity and distinguishing ‘indi-
viduals that stay’ from ‘individuals that move’, this model
circumvents the computational issues usually encountered
in standard multisite CR models when the number of sites
is large. Lagrange’s model is based on states that embed
information on whether an individual at t occupies the same
site as it occupied at t − 1 (‘S’ for ‘stay’) or not (‘M’ for
‘move’), as well as information about whether the individual
was captured (‘+’) or not (‘o’) at t − 1 and at t. We adapted
this parameterization to estimate the proportion of adult dis-
persers (i.e. the sexually mature individuals that changed lek
at least once during their lifetime) per generation. The dura-
tion of the CR survey corresponded to approximately two
western capercaillie generations in our population; the mean
adult life expectancy derived from our survival estimates
(see ‘Results’) was 3.16 years [ϕ/(1− ϕ), where ϕ is the sur-
vival probability]. Therefore, we can reasonably assume that
the proportion of dispersers estimated in our model corre-
sponds to the breeding dispersal rate per generation. Note
that we did not separate natal and breeding dispersal in our
model as individual age cannot be assessed in our study sys-
tem. Hence, fully resident individuals were assumed to be
strictly philopatric; in contrast, dispersers may have moved
before and/or after sexual maturity. This required extending
Lagrange’s model to consider additional states that include
4
information about fully resident behaviour (‘R’ for ‘fully
resident’). In our model, an individual could belong to two
alternative dispersal strategies during its lifetime: either a
disperser that may be in state ‘S’ or ‘M’ depending on its
movement status between t − 1 and t; or a fully resident indi-
vidual that always remained in the same lek (‘R’). This led
to the consideration of 10 states and 4 events (Appendix
B, Table B1 in ESM), which were coded in an individual’s
capture history and reflect the information available to the
observer at the time of capture (Fig. 2).
To quantify the proportion of dispersers per genera-
tion, we coded the initial states of the model in two steps
(Fig. 2): the first step embedded the probability μ that an
individual is assigned to the dispersing strategy; 1− μ cor-
responds to a probability of belonging the fully resident
strategy. The second step included the movement and cap-
ture state of a newly encountered individual (Fig. 2). When
captured for the first time, a disperser could be in state
oS + or oM + . Yet as these two states cannot be distin-
guished in Lagrange’s model or its extensions (Lagrange
et al. 2014; Cayuela et al. 2017, 2018a, b), we arbitrar-
ily kept the state oM + (Fig. 2). Fully resident individu-
als were assigned the state oR + . We then considered
three modelling steps in which the information of the
state descriptor was progressively updated in the model:
survival (ϕ), dispersal (ψ) and recapture (p). Each step
was conditional on all previous steps. As is standard in
a multievent model framework (Pradel 2005; Lagrange
et al. 2014), in the transition matrix, the rows correspond
to time t − 1, the columns to time t, and whenever a sta-
tus element is updated to its situation at t, it is shown in
http://doc.rero.ch
Fig. 2 The CR multievent model with its state-transition and event
matrices for quantifying the proportion of non-effective dispersers per
generation. The initial states were coded in two steps (initial states 1
and 2). In initial state 1: μ = proportion of dispersers per generation,
E = dispersers, R = fully resident individuals. The state–state transi-
tions were then modelled through three steps: survival ϕ, dispersal ψ,
bold and stays bold throughout the following steps. In the
first step, survival information was updated: an individual
could survive with a probability of ϕ or could die (d) with
a probability of 1− ϕ. This resulted in a transition matrix
with 10 states of departure and 7 intermediate states of
arrival (Fig. 2). In the second modelling step, dispersal
was updated. An individual belonging to the dispersing
strategy could move (M) from the lek it occupied with
a probability of ψ or could stay (S) with a probability of
1− ψ. An individual belonging the fully resident strategy
always stayed in the same lek; its probability of staying in
the same lek was fixed at 1. This led to a matrix of seven
states of departure and seven states of arrival (Fig. 2). In
the third step, the recapture information was updated: an
individual could be recaptured with a probability of p or
not with a probability of 1− p.
This parameterization was implemented in the E-surge
program (Choquet et al. 2009b). Competing models were
ranked through a model-selection procedure using Akaike
information criteria adjusted for a small sample size (AICc)
and AICc weights. If the AICc weight of the best-supported
model was less than 0.90, we performed a model-averaging
procedure. Our hypotheses concerning survival, disper-
sal and recapture were examined using the general model
[μ(SEX), ϕ(SEX), ψ(SEX), p(SEX + YEAR)], in which two
effects were considered: a sex effect (SEX), included in the
model as a discrete covariate; and a year effect (YEAR).
From this general model, we tested all possible combina-
tions of effects and ran 32 competing models (Appendix C,
Table C1 in ESM).
5
and recapture p. In the transition matrix, the rows correspond to time
t − 1, the columns to time t; whenever a status element is updated to
its situation at t, it is shown in bold and stays bold throughout the
following steps. For a detailed description of states and events, see
Appendix B, Table B1 in ESM
Quantifying annual non-effective dispersal rates
and distances
In a recent paper, Tournier et al. (2017) extended Lagrange’s
model by breaking down dispersal (ψ) into the distinct
parameters of departure (τ) and arrival (α). This new param-
eterization allows the quantification of the proportion of
individuals arriving in sites with different characteristics
or located at different distances from the source site. We
adapted this parameterization for the needs of our study to
consider states incorporating information about the capture
of an individual (‘+’ or ‘o’) at t − 1 and t and its movement
status. We also included information about the Euclidian
distance covered by dispersers between a departure and
an arrival patch. This was incorporated in the model using
five Euclidean distance classes (noted ‘1’ to ‘5’)—to date,
Euclidean distance cannot be included as a continuous vari-
able in multistate and multievent capture-recapture models.
The distance classes were fixed according to the movement
distances recorded in the population, ranging from less than
1 km to 29 km: 0–6 km (‘1’), 6–12 km (‘2’), 12–18 km (‘3’),
18–24 km (‘4’) and 24–30 km (‘5’). The number of dis-
tances classes was limited to five to facilitate model imple-
mentation; the number of model states increases with the
number of classes. This led to the consideration of 19 states
and 8 events (Appendix B, Table B2 in ESM).
When captured for the first time, the state of an individual
could be oS + or oM + . Yet as these two states cannot be dis-
tinguished in Lagrange’s model or its extensions (Lagrange
et al. 2014; Cayuela et al. 2017), we arbitrarily kept the state
 oS + (Fig. 3). We then considered four modelling steps: sur-
vival (ϕ), departure (τ), arrival (α), and recapture (p). In the
first step, the information about survival was updated. An
individual could survive with a probability of ϕ or could
die (d) with a probability of 1− ϕ. This led to a matrix with
19 states of departure and 5 intermediate states of arrival
(Fig. 3). In the second step, departure was updated. An
individual could move (M) from the lek it occupied with
a probability of τ or could stay (S) with a probability of
1− τ. This led to a matrix of five states of departure and five
states of arrival (Fig. 3). In the third step, we updated the
information about arrival. An individual that moved could
arrive in a patch located in the first four distance classes (1,
http://doc.rero.ch
Fig. 3 The CR multievent model with its state-transition and event
matrices for quantifying annual non-effective dispersal rates and dis-
tances. The initial states were coded in a single step. The state–state
transitions were then modelled through four steps: survival ϕ, depar-
ture τ, arrival α and recapture p. In the transition matrix, the rows
6
2, 3 or 4) from the source patch with a probability of α1, α2,
α3 or α4, or could arrive in a lek located in distance class
5 with a probability of 1–(α1 + α2 + α3 + α4). This led to
a matrix with 5 states of departure and 13 states of arrival
(Fig. 3). In the fourth step, recapture was updated (Fig. 3).
An individual could be recaptured with a probability of p
or not with a probability of 1− p, resulting in a transition
matrix with 13 states of departure and 19 states of arrival.
The last component of the model linked events to states. In
this specific situation, each state corresponded to only one
possible event (Fig. 3). The information about individuals’
dispersal status, distance, and capture is coded in the events
(Appendix B, Table B2 in ESM).
correspond to time t − 1, the columns to time t, and whenever a status
element is updated to its situation at t, it is shown in bold and stays
bold throughout the following steps. For a detailed description of
states and events, see Appendix B, Table B2 in ESM
 As for the previous model, this parameterization was
implemented in the E-surge program. Competing models
were ranked through a model-selection procedure (provided
in Appendix C in ESM) using Akaike information criteria
adjusted for a small sample size (AICc) and AICc weights.
We performed model averaging if the AICc weight of the
best-supported model was less than 0.90. Our hypotheses
concerning state–state transition probabilities were tested
using the general model [ϕ(SEX), τ(SEX), α(5 + SEX),
p(SEX + YEAR)]. We hypothesized that survival (ϕ) and
departure (τ) probabilities depended on sex (SEX). We also
examined whether arrival probability (α) differed among dis-
tance classes (5) and sex (SEX). Furthermore, we hypoth-
esized that recapture probability (p) depended on sex (SEX)
and varied among years (YEAR). From the general model,
we tested all the possible combinations of effects and built
32 competitive models (Appendix C, Table C1 in ESM).
http://doc.rero.ch
Assessing spatial genetic structure and relatedness
We examined the spatial genetic structure and relatedness for
both sexes. A total of 92 individuals were kept for the popu-
lation genetics analyses. We used 11 microsatellite markers
for our analyses (ADL184, ADL230, BG15, BG16, BG18,
LEI098, TuT1, TuT2, TuT3, TuT4, ADL142; see Appendix
A in ESM). We used genetic approaches designed to detect
contemporary genetic structure and relatedness pattern. The
Bayesian clustering approach (STRUCTURE) used in our
analyses is designed to detect genetic clusters in natural pop-
ulations (Pritchard et al. 2000; Corander et al. 2003; Broquet
and Petit 2009). Parentage analyses typically describe the
genealogical relationships (i.e. relatedness level) between
individuals through successive generation and applies to
recent times (Wang 2004, 2012; Städele and Vigilant 2016).
Furthermore, the use of individual-based approaches show
less inertia than approaches (e.g. Fst-based methods) assess-
ing the evolution of allele frequencies through many genera-
tions and are therefore more suitable to analyse contempo-
rary gene flow (Landguth et al. 2010). Hence, we can assume
that non-effective dispersal, relatedness structure and gene
flow are inferred at relatively similar (and recent) time scale.
Note that we were not able to investigate time-specific vari-
ation of the genetic structure because of the small size of
population; too few males and females were available each
year to perform robust analyses. As in most of genetic stud-
ies, we assumed that individuals belong to the lek where
they were detected for the first time during the study period.
Examining genetic structure
Evidence for scoring errors, large allele dropout, presence
of null alleles, departure from Hardy–Weinberg equilib-
rium (HWE) and linkage disequilibrium (LD) between all
7
pairs of loci were assessed with Micro-Checker v2.2.3 and
ARLEQUIN v3.5.1.3 (Van Oosterhout et al. 2004, Girard
and Angers 2008; Excoffier et al. 2005). These tests were
conducted only on the four large leks containing at least ten
individuals (NOI = 10, LLC = 12, VEN = 20 and GdF = 26)
with the Bonferroni correction for multiple comparisons
(p < 0.05; Rice 1989). Analyses of the spatial genetic struc-
ture were performed for all individuals, for males only and
for females only with Bayesian clustering of genotypic
data, implemented in STRUCTURE v2.3.3 (Pritchard et al.
2000), to estimate the most likely number of homogene-
ous genetic clusters (Appendix D in ESM). The Markov
chain Monte Carlo (MCMC) algorithm was run using the
admixture model with correlated allele frequencies (Falush
et al. 2003) for 500,000 steps after 2,000,000 initial burn-in
steps, with no a priori information on an individual’s lek.
Ten independent runs were performed for each value of K,
from 1 to 10. The most likely number of genetic clusters in
the dataset was suggested with the deltaK method (Evanno
et al. 2005), which was applied using tools available from
the STRU CTU RE HARVESTER website (Earl and von
Holdt 2012). However, the deltaK method is not appropri-
ate when the true K = 1, so we first tested for this possibility
by examining whether lnP(D) (an estimate of the posterior
probability of the data for a given K) was maximum for
K = 1. We also verified that the different runs for each K
were similar using the similarity coefficients (H’ values) of
the software CLUMPP v 1.1.1 (Jakobsson and Rosenberg
2007). The Greedy algorithm was used for K = 1–5, then the
LargeKGreedy algorithm was used for K = 6–10. To confirm
that sex-specific genetic structure (see ‘Results’) was linked
to higher genetic proximity of males within leks, we first
compared male and female genetic structure with a discri-
minant analysis of principal components (DAPC) using leks
as a priori (DAPC; Jombart et al. 2010). The DAPCs were
computed with the adegenet package in R (Jombart 2008).
Examining relatedness structure
We examined relatedness structure by comparing intra-
and inter-lek relatedness (rxy) for males and females using
COANCESTRY v1.0.1.8 (Wang 2011). Simulations were
performed to identify the best relatedness estimator; these
consisted of 600 dyads spread equally across six categories
of relatedness: parent–offspring (rxy = 0.5), full siblings
(rxy = 0.5), half siblings/avuncular/grandparent–grandchild
(rxy = 0.25), first cousins (rxy = 0.125), second cousins
(rxy = 0.03125), and unrelated (rxy = 0). In addition, we
used a permutation approach to test if the average related-
ness of an individual was significantly higher within the lek
of its first capture than a randomly chosen lek. This per-
mutation approach was performed separately for males and
females in the goal of comparing pattern between sex. We
 first computed the average relatedness of all individuals to
each lek (based on the males present in the lek and exclud-
ing the individual if present). For each individual, we then
randomly chose a lek and created a vector of average related-
ness of the permutated ‘first capture’ lek. The mean of this
vector was calculated, and a distribution of this statistic was
drawn based on 1000 iterations and compared with what we
observed in our original dataset. The p value was calculated
by dividing the number of values in the distribution that
were higher or equal to the original dataset with the number
of permutations (i.e. 1000). The same approach was then
used with departure from a lek to test if average relatedness
might have an impact on the decision of an individual to
emigrate from a given lek and compare pattern between sex.
Estimating effective dispersal from genetic data
http://doc.rero.ch
We assessed effective dispersal (i.e. gene flow) distance in
males and females while considering potential effects of
landscape features (topographic distance, slope and ridge)
on genetic differentiation in our analyses.
Inter-individual measures of genetic differentiation
The use of autosomal nuclear markers does not prevent the
detection of sex-specific differences in dispersal as long as
only adults are sampled (Goudet et al. 2002). All the genetic
analyses were thus performed separately on males (dataset
M) and females (dataset F). For each dataset, we computed
three kinds of inter-individual pairwise genetic distances: a
measure rxy of genetic dissimilarity computed as 1 − rxy,
the Bray–Curtis percentage dissimilarity metric (Bc; Leg-
endre and Legendre 1998) and the dissimilarity metric based
on the proportion of shared alleles at each locus (Dps). In
each dataset, the three metrics were highly correlated (Pear-
son’s correlation coefficient > 0.70).
Statistical analyses
For each dataset (M and F) and each measure of genetic dif-
ferentiation GD (rxy, Bc and Dps), we first performed a spa-
tial autocorrelation analysis with a non-directional Mantel
correlogram (Smouse and Peakall 1999) using the MATLAB
software-coding environment (Mathworks, Inc.) to deter-
mine the spatial scale of indirect gene flow (Anderson et al.
2010). For this purpose, Euclidean distance classes were
defined every 5000 m (up to 40 km), resulting in 8 binary
matrices representing the membership of individual pairs
to the distance class tested (with ‘0’ for pairs of individu-
als belonging to the same distance class and ‘1’ otherwise).
Each binary matrix was compared to the rxy, Bc or the Dps
matrix using a simple Mantel test with 1000 permutations.
We then plotted Mantel correlation values over distance
8
classes, with a 95% confidence interval determined by the
random removal of 20% of individuals (1000 iterations;
Peterman et al. 2014).
We also evaluated sex-specific variation in effective dis-
persal distance while controlling for functional landscape
connectivity. In addition to Euclidean distance (ED), we
considered three kinds of landscape predictors (see Appen-
dix E in ESM for details): topographic distance (Dtopo)
(assuming that individuals may not fly in a straight line,
but may follow the topographic relief), effective distance
along slopes (Dslope) (assuming that individuals may avoid
steep zones) and effective distance along ridges (Dridge)
(assuming that individuals may prefer ridge zones). For
each measure of genetic differentiation GD (rxy, Bc and
Dps) and for each sex, we designed a complete model as
follows: GD ∼ ED + Dtopo + Dslope + Dridge . We did not run
this model on all possible pairs of individuals, but selected
various subsets of pairwise data by defining a maximum
Euclidean distance of S between sample points. S ranged
from 14,000 m (a distance chosen so that no individual was
excluded from the neighbouring graph; Jombart et al. 2008)
to 46,000 m (the maximum Euclidean distance between two
individuals) in 1000 m increments. For each subset, we ran
the full model in a standard multiple linear regression and
plotted the model fit R2 against S to identify, for each sex, the
spatial scale that optimized the amount of variance explained
in each metric of genetic differentiation.
Results
Over the 5-year study period, we collected a total of 1341
samples of faeces and feathers (see detailed information in
Appendix A in ESM) assigned to 132 individuals (70 males,
59 females and three individuals for which sex could not be
identified). We kept in our annual capture–recapture data
a total of 236 observations. We detected 27 between-leks
dispersal events, 13 and 14 dispersal events in males and
females respectively.
Estimating non-effective dispersal
Concerning the proportion of dispersers per genera-
tion, the best-supported CR model was[μ (.), ϕ(.), ψ(.),
p(sex)] (Appendix C, Table C1 in ESM). However, as the
AICc difference with the second-ranked model was small
(ΔAICc = 0.41), we performed model averaging (from the
complete set of models). Model-averaged estimates of recap-
ture probabilities indicated that males (p = 0.72 ± 0.06) were
more detectable than females (p = 0.46 ± 0.06). Survival did
not differ substantially between the two sexes; the survival
probability of males (ϕ = 0.75 ± 0.04) was only marginally
lower than that of females (ϕ = 0.79 ± 0.06). The proportion
 of dispersers per generation (μ) was also relatively similar in
both sexes; 0.38 ± 0.09 in females and 0.30 ± 0.15 in males.
In the individuals with a dispersing strategy, the annual
dispersal probability ψ did not differ between sexes; it was
0.67 ± 0.15 in females and 0.64 ± 0.15 in males.
Concerning annual non-effective dispersal rates and
distances, the best-supported CR model was [ϕ(.), τ(sex),
α(5), p(sex)] (Appendix C, Table C2 in ESM). Again, as the
AICc difference with the second-ranked model was small
(ΔAICc = 0.05), we performed model averaging (from
the complete set of models). The annual departure rate
slightly differed between sexes: females had very margin-
ally higher dispersal probability (τ = 0.23 ± 0.15) than males
(τ = 0.16 ± 0.07). Note that departure rate showed little vari-
ation between years (see Appendix C in ESM). Arrival prob-
ability was similar between sexes and decreased with the
Euclidean distance between leks (Fig. 4). The probability of
http://doc.rero.ch
arriving in a lek located at a distance ranging from 0 to 6 km
was 0.49 ± 0.12 in females and 0.50 ± 0.12 in males. At dis-
tances between 6 and 24 km, the arrival probability was rela-
tively constant (ranging from 0.11 to 0.17) in both sexes. At
distances from 24 to 30 km, arrival probability substantially
decreased (0.03 ± 0.03 in females and 0.07 ± 0.07 in males).
Fig. 4 Annual non-effective dispersal distance in T. urogallus [left:
females (n = 48), right: males (n = 61)]. The arrival probabilities (with
standard errors shown in error bars) were model averaged
9
Assessing spatial genetic structure and relatedness
We found no evidence of genotyping errors, large allele
dropout or null alleles. No significant deviation from the
Hardy–Weinberg equilibrium (HWE) was observed in
three of the four large leks; linkage disequilibrium (LD)
was observed between the loci ADL230 and BG15 for the
lek NOI after Bonferonni correction (p < 0.05). Because the
small sample size could be linked to this LD, we kept these
loci for further analyses and thus consider our molecular
marker dataset as neutral and biparentally inherited. Miss-
ing data correspond to 0.3% of the final dataset and were
considered negligible.
When considering all individuals, the optimal number of
genetic clusters found with the STRUCTURE software was
two (Appendix D, Fig. D1 in ESM). However, the optimal
number of genetic clusters for males alone was three, and for
females it was one (Fig. 5; Appendix D, Fig. D2 and D3 in
ESM). The DAPCs suggest higher male genetic proximity
inside a given lek in comparison to females (Appendix D,
Fig. D4 and D5 in ESM).
Concerning relatedness analyses, according to our simu-
lations, the best rxy estimator was LynchRD, which has a
similar mean rxy in comparison to true values (0.22 versus
0.23) and a correlation of 0.61 (Appendix D, Table D1 in
ESM). Based on this estimator, males and females differed in
terms of relatedness. Males had a significantly higher intra-
lek average relatedness than females (t = 5.10, df = 254.55,
p < 0.001; Fig. 6). Both males and females had significantly
higher average relatedness within the lek of their first cap-
ture (p < 0.001 for both, see Fig. 6), but only females showed
significantly higher average relatedness in leks from which
they emigrated (p = 0.042; Appendix F, Fig. F1 in ESM).
Estimating effective dispersal distance
Autocorrelograms based on rxy indicated spatial patterns of
isolation by distance in both sexes, with significant positive
spatial autocorrelation occurring up to 10 km in both males
and females (Fig.  7a, b). Nevertheless, females showed
positive spatial autocorrelation up to 15 km, whereas males
showed significant negative autocorrelation from 15 km,
suggesting higher effective dispersal distances in females
than in males. Patterns of isolation by distance were very
similar when considering Bc and Dps genetic distances.
The plot of model fit R 2 against maximum pairwise
Euclidean distances indicated that model fit was always
higher in males (R2 ranging from 0.115 to 0.355) than in
females (R2 ranging from 0.012 to 0.076), whatever the
Euclidean distance and the considered measure of genetic
differentiation (Fig. 7c). This did not result from unbal-
anced sample size in males (n = 51) and females (n = 41).
Furthermore, model fit values were always higher (or at
http://doc.rero.ch
Fig. 5 Outputs from STRUCTURE software showing high population genetic structure in males (a K = 3) and the absence of genetic structure in
females (b K = 1). The sample size was n = 51 in males and n = 41 in females
least equal) when using rxy rather than Bc or Dps, in both
sexes. In males, the optimal maximum Euclidean distance
was 15,000 m (R 2 = 0.355), whereas it was 36,000 m in
females (R2 = 0.076). In both sexes, the main contributor to
the variance in measures of genetic differentiation was the
topographic distance, indicating that topographic relief is
a better predictor of effective dispersal than Euclidean dis-
tance in this species (see Appendix E in ESM for details).
Fig. 6 Spatial structure of relatedness in the study area population
of T. urogallus (top: males; bottom: females). The mean relatedness
coefficients r were calculated within leks (circles) and between leks
(arrows) and are provided with their 95% CI. The sample size was
n = 51 in males and n = 41 in females
10
Comparing non-effective and effective dispersal
patterns
Overall, our results showed that non-effective and effec-
tive dispersal patterns differ between sexes. In females, the
effective dispersal distances were highly congruent with the
non-effective dispersal distances. The proportion of female
dispersers settling in a lek located at a distance of more than
24 km was dramatically lower (< 5%) (Fig. 4), which is in
accordance with the upper estimate of gene flow distance
(26 km) (Fig. 7b). In males, the non-effective dispersal dis-
tances strongly exceeded effective dispersal distances. The
pattern of male non-effective dispersal distance was rela-
tively similar to that found in females (Fig. 4). Yet the maxi-
mum gene flow distance was 12 km (Fig. 7a), resulting in
a 12-km gap between non-effective and effective dispersal
distances in males.
Discussion
Our findings revealed stronger spatial structure of related-
ness in males than in females; however, males disperse at a
similar rate and distance than females. They also indicated
that the population viscosity could allow male cooperation
through two non-exclusive mechanisms. First, the results
http://doc.rero.ch
Fig. 7 Effective dispersal distances in T. urogallus. a and b Mantel
correlograms showing the relationships between rxy distances and
Euclidean distance classes (defined every 5000  m) in males (a) and
females (b). The dots indicate the value of r: standard Mantel cor-
relation with 1000 permutations (black circles, p value < 0.05). Error
bars bound the 95% CI of r as determined by the random removal of
20% of populations. Upper and lower confidence limits (dotted lines)
bound the 95% CI of the null hypothesis of no spatial structure as
showed that at their first lek attendance, males aggregated
in a lek composed of relatives and tended not to leave this
patch in order to avoid kin competition. Second, non-effec-
tive dispersal distance dramatically outweighed effective
dispersal distance, which suggests that dispersers incurred
high post-settlement costs. These two mechanisms result
in strong population genetic structure in males. In females,
our study revealed a lower spatial structure of relatedness
compared to males. At their first lek attendance, females
arrived in a lek composed of relatives. Yet they also pref-
erentially tended to emigrate from this type of lek, likely
due to inbreeding risk, which weakens the spatial structure
of relatedness. Our results also found that non-effective
dispersal and effective dispersal distances were highly
similar in females, which suggests the absence of post-
settlement costs. So both this and inbreeding avoidance
are likely to explain the low genetic structure in females.
11
determined by permutations. c Plot of model fit against maximum
pairwise Euclidean distance in males (black) and females (grey).
Solid lines: model fit with rxy dissimilarity metric as the dependent
variable. Dashed lines: model fit with Bc dissimilarity metric as the
dependent variable. Dotted lines: model fit with Dps dissimilarity
metric as the dependent variable. Arrows indicate the best combina-
tion of maximum Euclidean distance and genetic distance for each
sex. The sample size was 51 in males and n = 41 in females
Non-effective dispersal rate is high and does
not differ between sexes
Our study revealed that non-effective dispersal rates per gen-
eration and annual dispersal probability were relatively simi-
lar for both sexes. Overall, non-effective dispersal between
leks was high in our study area: 38% of females and 30%
of males dispersed at least once during their lifetime. The
annual non-effective dispersal rates were also high: 23% in
females and 16% in males. This result goes against the con-
ventional view that T. urogallus males are highly philopatric
(Storch 1997; Höglund and Alatalo 2014). Yet to our knowl-
edge, our study is the first to use long-term CR data to quan-
tify dispersal rates in this species. In addition, while previous
studies have mainly focused on movements at a relatively
small spatial scale (Hjeljord et al. 2000; Wegge et al. 2007),
the large spatial extent of our study allowed the detection
of long-distance dispersal events (max. = 29 km). Although
 our sample size was relatively small (n = 107 individuals),
we are confident that the patterns drawn in our study are reli-
able. First, our dataset is nearly exhaustive since all known
active leks from the Vosges Mountains and (almost) all indi-
viduals composing those leks have been considered in our
analyses. Therefore, we can rule out the possibility that the
dispersal pattern detected in our analyses result from weak-
nesses in the sampling design. Second, if any difference in
dispersal rate (or distance) actually exists between sexes,
it should be small. Furthermore, although the structure of
capture–recapture models used in this study is sophisticated,
the number of model parameters was relatively small; many
states are structural and are not estimated as they are not
biologically relevant. In addition, E-SURGE diagnostics
(Choquet et al. 2009a, b) indicated that all parameters were
estimable despite the small sample size. Therefore, we are
confident in the inferences of our models about breeding
http://doc.rero.ch
dispersal patterns. However, we were not able to investigate
natal dispersal—a central process in bird dispersal (Para-
dis et al. 1998)—since individuals could not be “captured”
before their first spring, i.e., after natal dispersal. We nev-
ertheless tried to investigate lek choice at (presumably) first
lek attendance in an indirect way via relatedness analyses.
Females likely avoid inbreeding and do not pay high
dispersal costs
Our results highlighted that the difference between intra-
lek and inter-lek relatedness was relatively weak in females.
Within leks, the mean coefficient of relatedness was 0.09
(SE 0.03), which corresponds to genealogical levels of first
and second cousins; between leks, the mean coefficient was
− 0.06 (SE 0.01), indicating that females are only weakly
related to the lekking males. The findings indicated that,
at their first detection in a lek, females arrived in a lek
composed of close relatives—as the recapture probability
of female is relatively low (≈ 0.50), it was not possible to
make the assumption that the first detection of a female cor-
responds to its first lek visit. Yet, our study revealed that 38%
of females change lek at least once during their lifetime; dis-
persing females have a 64% chance of dispersing each year.
We found that females preferentially emigrate from leks of
males that are close relatives (Appendix F, Fig. F1 in ESM),
likely to avoid inbreeding risks. A similar inbreeding-avoid-
ance mechanism was highlighted by Lebigre et al. (2010) in
the closely related species Tetrao tetrix. Furthermore, our
results showed that non-effective dispersal distances for
females (less than 5% of dispersers move farther than 24 km)
are highly congruent with effective dispersal distances (the
maximum gene flow distance is 26 km). This suggests the
absence of substantial reproductive costs after settling into
a new lek, likely due to reduced female–female competition
for mate acquisition in lekking birds (Höglund and Alatalo
12
2014). In females, inbreeding avoidance behaviour and the
absence of high post-settlement costs likely reduce the spa-
tial structure of relatedness, as well as the population genetic
structure.
Context-dependent dispersal and high
post-settlement costs may facilitate cooperative
courtship in males
In contrast, in males, the results revealed the existence of a
strong spatial structure of relatedness. Within leks, the mean
coefficient of relatedness was 0.25 (SE 0.02), which cor-
responds to genealogical levels equal to half sibling, avun-
cular or grandparent–grandchild. Between leks, the mean
coefficient was − 0.06 (SEM 0.01), indicating that males
are unrelated. This pattern is congruent with those found in
two previous studies on T. urogallus (Regnaut et al. 2006;
Segelbacher et al. 2007) and indicates that males join leks
composed of relatives. The pattern is more consistent with a
preference in lek choice than with a negative effect of land-
scape features on dispersal since our molecular analyses
showed genetic differentiation was lowly affected by land-
scape connectivity. Among the different landscape predictors
introduced in the regression analyses, only the topographic
distance had an influence on genetic differentiation, but the
strength of the effect was relatively small.
Our study revealed that this spatial structure of relat-
edness is maintained despite relatively high non-effective
dispersal: 30% of the males disperse at least once during
their lifetime, and these males have a high propensity to
disperse each year (0.65). These findings suggest that the
maintenance of male population viscosity is based on two
non-exclusive mechanisms. First, after brood dissolution,
young males tend to aggregate in leks composed of close
relatives—as the recapture probability of males is relatively
high (≈ 0.80), one can assume that the first detection of an
individual corresponds to its first attendance to a lek. When
they emigrate from a lek, emigration does not seem to be
associated with the social context within the lek of depar-
ture; permutation analyses indicate that relatedness does
affect emigration. The high capture probability of males
(0.72) does not suggest that we failed to detect an effect due
to a methodological bias. Yet it should be kept in mind that
the number of dispersal events was limited in our dataset,
which likely limited our ability to detect a signal. Second,
the findings suggest that male population viscosity might
also be maintained by high dispersal costs. Non-effective
dispersal distance dramatically outweighed effective dis-
persal distance, suggesting that dispersers may pay acute
post-settlement costs. The fitness loss after settling into
a new lek composed of non-relatives could be caused by
intense male–male competition, associated with territorial
and agonistic behaviour (Höglund and Alatalo 2014) and/
 or the loss of kin-selected benefits (Krakauer 2005; DuVal
2007). However, although these results are highly congru-
ent with predictions of theoretical models (El Mouden and
Gardner 2008), we are aware that the evidence is indirect.
First, we do not know how male inclusive fitness varies
according to the presence/number of relatives in a lek. Sec-
ond, lek size could affect lek choice by males (Kokko and
Lindstrom 1996), but the relative influence of kin-selected
benefits and lek size on male dispersal decisions are still
unknown in T. urogallus. Third, the type of cooperation is
poorly known in capercaillies and we do not have behav-
ioural data that directly show that relative capercaillie males
cooperate. Nonetheless, we can safely assume that it is closer
to a system based on alliances among relatives, as described
in wild turkeys (Krakauer 2005), than to a system with α and
β unrelated males as in manakins (DuVal 2007). Pedigree
analyses (which were not feasible in our study due to the
http://doc.rero.ch
low number of neutral markers) and experimental studies
should be undertaken to address these issues and confirm
the conclusions of our study.
In summary, our results suggest that male aggregation
in leks composed of relatives and enhanced dispersal costs
could maintain population viscosity, creating the conditions
conducive to cooperative courtship. While cooperative males
are more likely to benefit kin in viscous populations via
inclusive fitness, they must also compete for limited mating
opportunities with these same kin (Platt and Bever 2009).
As a result, kin competition could cancel out to some degree
the benefits of kin cooperation, possibly promoting kin-
dependent dispersal in lek systems. However, in our study,
we failed to detect an effect of kin competition on departure
probability. The strong spatial structure of relatedness indi-
cates that, if dispersal driven by kin competition exists, its
consequences on lek social composition is weak. This might
be due in part to the fact that the recent decline of the popu-
lation in our study area (possibly caused by anthropogenic
habitat degradation) has resulted in a marked decrease in lek
size, which may limit the risk of kin competition.
Role of direct and indirect benefits in lek system
evolution
Previous studies reported contradictory results about the role
kin selection in the evolution of bird lek systems. Several
studies in manakins (Loiselle et al. 2006; McDonald 2009;
Davis et al. 2015), bowerbirds (Madden et al. 2004) and
grouses (Gibson et al. 2005; Lebigre et al. 2008; Bush et al.
2010) did not detect any increase of relatedness level within
leks and concluded that the maintenance of male aggrega-
tions could mostly be due to direct rather than indirect fitness
benefits. Three alternative hypotheses have been proposed:
the hot-spot hypothesis states that males are hypothesized
to sequentially cluster in areas of high female density or
13
movement (Bradbury and Gibson 1983); the hotshot hypoth-
esis postulates that subordinate males settle near dominant
males with high reproductive success (Beehler and Foster
1988); in the delayed benefits hypothesis, subordinate males
receive direct-fitness benefits later in life when they replace
higher ranking males on the leks (McDonald and Potts 1994;
Kokko and Johnstone 1999). To date, the role of these dif-
ferent mechanisms in lek system evolution is still debated
(Loiselle et al. 2006; Höglund and Alatalo 2014). Other
studies, as ours, detected an increase of the relatedness level
within leks in manakins (Shorey et al. 2000; Höglund and
Shorey 2003; Francisco et al. 2009; Concannon et al. 2012),
bowerbirds (Reynolds et al. 2009), peafowls (Petrie et al.
1999), wild turkeys (Krakauer 2005), and grouses (Höglund
et al. 1999; Regnaut et al. 2006). These studied defended
the hypothesis that kin selection and indirect fitness benefits
play a central role in the evolution of cooperative breeding
systems. They stated that population viscosity can be main-
tained by male philopatry and kin-recognition mechanism
including self referent phenotypic matching (Petrie et al.
1999). In this context, our study also suggests that males
dispersing to leks composed of non-relatives may incur high
post-settlement costs due to the loss of kin-selected ben-
efits. To date, the role of direct and indirect benefits in male
aggregation and lek system formation still remains contro-
versial. Broader analyses should be undertaken to quantify
within-lek relatedness level in a greater number of bird spe-
cies, and population per species, to draw general conclusions
about the mechanisms shaping lek system evolution.
Acknowledgements The genetic monitoring of capercaillie in the
Vosges Mountains was funded by the Life + Project “Des Forêts pour
le Grand tetras”, by the Natura2000 network and by the regional
programme of the Capercaillie National Action Plan initiated by the
French Ministry of Environment. The project largely relied on the work
of volunteers who collected samples during the 6 years of the study:
Antoine Andre, Didier Arseguel, Samuel Audinot, Alix Badre, Etienne
Barbier, Dominique Becker, Bernard Binetruy, Frédéric Bocquenet,
Noémie Castaing, Sebastien Coulette, Stéphane Damervalle, Luc
Dauphin, Richard Delaunay, Lucile Demaret, Michel Despoulin, Lau-
rent Domergue, Vincent Drillon, Christian Dronneau, Fabien Dupont,
Arnaud Foltzer, Patrick Foltzer, Marc Gehin, Cyril Gerard, Maxime
Girardin, Remi Grandemange, Jean-Claude Gregy, Joaquim Hatton,
Thibaut Hingray, Thierry Hue, Arnaud Hurstel, Jean-Nöel Journot,
Fabien Kilque, Lydie Lallement, Christian Lamboley, Manuel Lembke,
Jean-Michel Letz, Vincent Lis, Olivier Marchand, Paul Massard, Yvan
Mougel, Michel Munier, Louis-Michel Nageleisen, Yvan Nicolas,
Christian Oberle, Pascal Perrotey-Doridant, Christian Philipps, Fran-
çois Rey-Demaneuf, Dorian Toussaint, Jean-Marie Triboulot, Bruno
Vaxelaire, Laurent Verard, Jean-Lou Zimmermann, and Alice Zimmer-
mann. We also thank Jacob Höglund and the other anonymous referee
for their constructive comments that helped to improve the manuscript.
Author contribution statement GJ, JC and HC conceived the ideas and
designed methodology. FP and AL collected the data. HC, ML, FF,
JGP and LB analysed the data. HC led the writing of the manuscript.
All authors contributed critically to the drafts and gave final approval
for publication.
 References Corander J, Waldmann P, Sillanpää MJ (2003) Bayesian analy-
sis of genetic differentiation between populations. Genetics
163(1):367–374
Alatalo RV, Höglund J, Lundberg A, Sutherland WJ (1992) Evolution
Cotto O, Kubisch A, Ronce O (2014) Optimal life-history strategy
of black grouse leks: female preferences benefit males in larger
differs between philopatric and dispersing individuals in a meta-
leks. Behav Ecol 3:53–59
population. Am Nat 183:384–393
Anderson CD, Epperson BK, Fortin MJ, Holderegger R, James P,
Davis DM, Reese KP, Gardner SC, Bird KL (2015) Genetic struc-
Rosenberg MS, Scribner KT, Spear S (2010) Considering spatial
ture of Greater Sage-Grouse (Centrocercus urophasianus) in a
and temporal scale in landscape-genetic studies of gene flow. Mol
declining, peripheral population. Condor 117:530–544
Ecol 19:3565–3575
Dickinson JL, Euaparadorn M, Greenwald K, Mitra C, Shizuka D
Baguette M, Blanchet S, Legrand D, Stevens VM, Turlure C (2013)
(2009) Cooperation and competition: nepotistic tolerance and
Individual dispersal, landscape connectivity and ecological net-
intrasexual aggression in western bluebird winter groups. Anim
works. Biol Rev 88:310–326
Behav 77:867–872
Beehler BM, Foster MS (1988) Hotshots, hotspots, and female
DuVal EH (2007) Adaptive advantages of cooperative courtship for
preference in the organization of lek mating systems. Am Nat
subordinate male lance-tailed manakins. Am Nat 169:423–432
131:203–219
Earl DA, von Holdt BM (2012) STRU CTU RE HARVESTER: a
Bonte D, Van Dyck H, Bullock JM, Coulon A, Delgado M, Gibbs M,
website and program for visualizing STRU CTU RE output
Lehouck V, Matthysen E, Mustin K, Saastamoinen M, Schtick-
and implementing the Evanno method. Conserv Genet Resour
zelle N, Stevens VM, Vandewoestijne S, Baguette M, Barton K,
4:359–361
Benton TG, Chaput-Bardy A, Clobert J, Dytham C, Hovestadt T,
El Mouden C, Gardner A (2008) Nice natives and mean migrants:
Meier CM, Palmer SCF, Turlure C, Travis JMJ (2012) Costs of
the evolution of dispersal-dependent social behaviour in viscous
dispersal. Biol Rev 87:290–312
http://doc.rero.ch

populations. J Evol Biol 21:1480–1491

Bourke AF (2011) The validity and value of inclusive fitness theory.
Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clus-
Proc R Soc Lond B Biol Sci 278:3313–3320
ters of individuals using the software STRUCTURE: a simulation
Bradbury JW, Gibson RM (1983) Leks and mate choice. In: Batson
study. Mol Ecol 14:2611–2620
P (ed) mate choice. Cambridge University Press, Cambridge, pp
Excoffier L, Laval G, Schneider S (2005) Arlequin ver. 3.0: an inte-
109–138
grated software package for population genetics data analysis.
Broquet T, Petit EJ (2009) Molecular estimation of dispersal for
Evol Bioinform 1:47–50
ecology and population genetics. Annu Rev Ecol Evol Syst
Falush D, Stephens M, Pritchard JK (2003) Inference of population
40:193–216
structure: extensions to linked loci and correlated allele frequen-
Bush KL, Aldridge CL, Carpenter JE, Paszkowski CA, Boyce MS,
cies. Genetics 164:1567–1587
Coltman DW, Scribner KT (2010) Birds of a feather do not always
Forero MG, Donázar JA, Hiraldo F (2002) Causes and fitness conse-
lek together: genetic diversity and kinship structure of greater
quences of natal dispersal in a population of black kites. Ecology
sage-grouse (Centrocercus urophasianus) in Alberta. Auk
83:858–872
127:343–353
Francisco MR, Gibbs HL, Galetti PM Jr (2009) Patterns of individual
Cayuela H, Pradel R, Joly P, Besnard A (2017) Analysing movement
relatedness at blue manakin (Chiroxiphia caudata) leks. Auk
behaviour and dynamic space-use strategies among habitats using
126:47–53
multi-event capture-recapture modelling. Methods Ecol Evol
Gibson RM, Pires D, Delaney KS, Wayne RK (2005) Microsatellite
8:1124–1132
DNA analysis shows that greater sage grouse leks are not kin
Cayuela H, Pradel R, Joly P, Bonnaire E, Besnard A (2018a) Estimat-
groups. Mol Ecol 14:4453–4459
ing dispersal in spatiotemporally variable environments using
Girard P, Angers B (2008) Assessment of power and accuracy of
multievent capture–recapture modeling. Ecology 99:1150–1163
methods for detection and frequency estimation of null alleles.
Cayuela H, Rougemont Q, Prunier JG, Moore JS, Clobert J, Besnard
Genetica 134:187–197
A, Bernatchez L (2018b) Demographic and genetic approaches
Goudet J, Perrin N, Waser P (2002) Tests for sex-biased dispersal using
to study dispersal in wild animal populations: a methodological
biparentally inherited genetic markers. Mol Ecol 11:1103–1114
review. Mol Ecol 27:3976–4010
Greenwood PJ, Harvey PH (1982) The natal and breeding dispersal of
Choquet R, Lebreton JD, Gimenez O, Reboulet AM, Pradel R (2009a)
birds. Annu Rev Ecol Syst 13:1–21
U-CARE: utilities for performing goodness of fit tests and manip-
Griffin AS, West SA (2003) Kin discrimination and the benefit of help-
ulating CApture–REcapture data. Ecography 32:1071–1074
ing in cooperatively breeding vertebrates. Science 302:634–636
Choquet R, Rouan L, Pradel R (2009b) Program E-SURGE: a software
Hamilton WD (1964) The genetical evolution of social behaviour. II.
application for fitting multievent models. In: Thomson DL, Cooch
J Theor Biol 7:17–52
EG, Conroy MJ (eds) Modeling demographic processes in marked
Hansson B, Bensch S, Hasselquist D (2004) Lifetime fitness of short-
populations. Springer, New York, pp 845–865
and long-distance dispersing great reed warblers. Evolution
Clarke AL, Sæther BE, Røskaft E (1997) Sex biases in avian dispersal:
58:2546–2557
a reappraisal. Oikos 79:429–438
Hepper PG (ed) (2005) Kin recognition. Cambridge University Press,
Clobert J, Galliard L, Cote J, Meylan S, Massot M (2009) Informed
Cambridge
dispersal, heterogeneity in animal dispersal syndromes and
Hjeljord O, Wegge P, Rolstad J, Ivanova M, Beshkarev AB (2000)
the dynamics of spatially structured populations. Ecol Lett
Spring-summer movements of male capercaillie Tetrao urogallus:
12:197–209
a test of the ‘landscape mosaic’ hypothesis. Wildl Biol 6:251–256
Clutton-Brock T (2002) Breeding together: kin selection and mutualism
Höglund J (2003) Lek-kin in birds—provoking theory and surprising
in cooperative vertebrates. Science 296:69–72
new results. Ann Zool Fenn 40:249–253
Clutton-Brock T (2009) Cooperation between non-kin in animal socie-
Höglund J, Alatalo RV (2014) Leks. Princeton University Press,
ties. Nature 462(7269):51
Princeton
Concannon MR, Stein AC, Uy JAC (2012) Kin selection may con-
Höglund J, Shorey L (2003) Local genetic structure in a white-bearded
tribute to lek evolution and trait introgression across an avian
manakin population. Mol Ecol 12(9):2457–2463
hybrid zone. Mol Ecol 21:1477–1486

14
 Höglund J, Alatalo RV, Lundberg A, Rintamäki PT, Lindell J (1999)
Microsatellite markers reveal the potential for kin selection on
black grouse leks. Proc R Soc Lond B Biol Sci 266:813–816
Jakobsson M, Rosenberg NA (2007) CLUMPP: a cluster matching
and permutation program for dealing with label switching and
multimodality in analysis of population structure. Bioinformatics
23:1801–1806
Jombart T (2008) adegenet: a R package for the multivariate analysis
of genetic markers. Bioinformatics 24:1403–1405
Jombart T, Devillard S, Dufour AB, Pontier D (2008) Revealing cryp-
tic spatial patterns in genetic variability by a new multivariate
method. Heredity 101:92–103
Jombart T, Devillard S, Balloux F (2010) Discriminant analysis of
principal components: a new method for the analysis of geneti-
cally structured populations. BMC Genet 11:94
Kokko H, Johnstone RA (1999) Social queuing in animal societies: a
dynamic model of reproductive skew. Proc R Soc Lond B Biol
Sci 266:571–578
Kokko H, Lindstrom J (1996) Kin selection and the evolution of leks:
whose success do young males maximize? Proc R Soc Lond B
Biol Sci 263:919–923
http://doc.rero.ch
Kokko H, Mackenzie A, Reynolds JD, Lindström J, Sutherland WJ
(1999) Measures of inequality are not equal. Am Nat 154:358–382
Komdeur J, Richardson DS, Hatchwell B (2008) Kin-recognition mech-
anisms in cooperative breeding systems: ecological causes and
behavioral consequences of variation. Ecology of social evolution.
Springer, Berlin, Heidelberg, pp 175–193
Krakauer AH (2005) Kin selection and cooperative courtship in wild
turkeys. Nature 434:69–72
Lagrange P, Pradel R, Bélisle M, Gimenez O (2014) Estimating disper-
sal among numerous sites using capture–recapture data. Ecology
95:2316–2323
Landguth EL, Cushman SA, Schwartz MK, McKelvey KS, Murphy
M, Luikart G (2010) Quantifying the lag time to detect barriers
in landscape genetics. Mol Ecol 19(19):4179–4191
Le Galliard JF, Ferrière R, Dieckmann U (2003) The adaptive dynam-
ics of altruism in spatially heterogeneous populations. Evolution
57:1–17
Le Galliard JF, Ferriere R, Dieckmann U (2004) Adaptive evolution of
social traits: origin, trajectories, and correlations of altruism and
mobility. Am Nat 165:206–224
Lebigre C, Alatalo RV, Forss HE, Siitari H (2008) Low levels of relat-
edness on black grouse leks despite male philopatry. Mol Ecol
17:4512–4521
Lebigre C, Alatalo RV, Siitari H (2010) Female-biased dispersal alone
can reduce the occurrence of inbreeding in black grouse (Tetrao
tetrix). Mol Ecol 19:1929–1939
Lebreton JD, Nichols JD, Barker RJ, Pradel R, Spendelow JA (2009)
Modeling individual animal histories with multistate capture–
recapture models. Adv Ecol Res 41:87–173
Legendre P, Legendre LFJ (1998) Numerical ecology. Elsevier Science
B.V, Amsterdam
Loiselle BA, Ryder TB, Duraes R, Tori W, Blake JG, Parker PG
(2006) Kin selection does not explain male aggregation at leks of
4 manakin species. Behav Ecol 18:287–291
Lowe WH, Allendorf FW (2010) What can genetics tell us about popu-
lation connectivity? Mol Ecol 19:3038–3051
Madden JR, Lowe TJ, Fuller HV, Coe RL, Dasmahapatra KK, Amos
W, Jury F (2004) Neighbouring male spotted bowerbirds are not
related, but do maraud each other. Anim Behav 68:751–758
Matthysen E (2012) Multicausality of dispersal: a review. Dispers Ecol
Evol 27:3–18
McDonald DB (2009) Young-boy networks without kin clusters in a
lek-mating manakin. Behav Ecol Sociobiol 63:1029–1034
McDonald DB, Potts WK (1994) Cooperative display and related-
ness among males in a lek-mating bird. Science 266:1030–1032
15
Møller AP, Alatalo RV (1999) Good-genes effects in sexual selec-
tion. Proc R Soc Lond B Biol Sci 266:85–91
Nystrand M (2007) Associating with kin affects the trade-off between
energy intake and exposure to predators in a social bird species.
Anim Behav 74:497–506
Paradis E, Baillie SR, Sutherland WJ, Gregory RD (1998) Pat-
terns of natal and breeding dispersal in birds. J Anim Ecol
67(4):518–536
Peterman WE, Connette GM, Semlitsch RD, Eggert LS (2014) Eco-
logical resistance surfaces predict fine-scale genetic differentiation
in a terrestrial woodland salamander. Mol Ecol 23:2402–2413
Petrie M, Krupa A, Burke T (1999) Peacocks lek with relatives even
in the absence of social and environmental cues. Nature 401:155
Piertney SB, Lambin X, Maccoll AD, Lock K, Bacon PJ, Dallas JF,
Leckie F, Mougeot F, Racey PA, Redpath S, Moss R (2008) Tem-
poral changes in kin structure through a population cycle in a ter-
ritorial bird, the red grouse Lagopus lagopus scoticus. Mol Ecol
17:2544–2551
Platt TG, Bever JD (2009) Kin competition and the evolution of coop-
eration. Trends Ecol Evol 24:370–377
Pradel R (2005) Multievent: an extension of multistate capture–recap-
ture models to uncertain states. Biometrics 61:442–447
Pritchard JK, Stephens M, Donnelly P (2000) Inference of population
structure using multilocus genotype data. Genetics 155:945–959
Prugnolle F, De Meeûs T (2002) Inferring sex-biased dispersal from
population genetic tools: a review. Heredity 88:161
Pusey AE (1987) Sex-biased dispersal and inbreeding avoidance in
birds and mammals. Trends Ecol Evol 2:295–299
Queller DC (1992) A general model for kin selection. Evolution
46:376–380
Regnaut S, Christe P, Chapuisat M, Fumagalli L (2006) Genotyping
faeces reveals facultative kin association on capercaillie’s leks.
Conserv Genet 7:665–674
Reynolds SM, Christman MC, Uy JAC, Patricelli GL, Braun MJ, Bor-
gia G (2009) Lekking satin bowerbird males aggregate with rela-
tives to mitigate aggression. Behav Ecol 20:410–415
Rice WR (1989) Analyzing tables of statistical tests. Evolution
43:223–225
Ronce O (2007) How does it feel to be like a rolling stone? Ten
questions about dispersal evolution. Annu Rev Ecol Evol Syst
38:231–253
Rosher C, Favati A, Dean R, Løvlie H (2017) Relatedness and age
reduce aggressive male interactions over mating in domestic fowl.
Behav Ecol 28:760–766
Segelbacher G, Wegge P, Sivkov AV, Höglund J (2007) Kin groups in
closely spaced capercaillie leks. J Ornithol 148:79–84
Shorey L, Piertney S, Stone J, Höglund J (2000) Fine-scale genetic
structuring on Manacus manacus leks. Nature 408:352
Sinervo B, Clobert J (2003) Morphs, dispersal behavior, genetic simi-
larity, and the evolution of cooperation. Science 300:1949–1951
Smouse PE, Peakall R (1999) Spatial autocorrelation analysis of
individual multiallele and multilocus genetic structure. Heredity
82:561–573
Städele V, Vigilant L (2016) Strategies for determining kinship in wild
populations using genetic data. Ecol Evol 6(17):6107–6120
Storch I (1997) Male territoriality, female range use, and spatial
organisation of capercaillie Tetrao urogallus leks. Wildl Biol
3(3/4):149–162
Szulkin M, Sheldon BC (2008) Dispersal as a means of inbreeding
avoidance in a wild bird population. Proc R Soc Lond B Biol Sci
275:703–711
Tournier E, Besnard A, Tournier V, Cayuela H (2017) Manipulating
waterbody hydroperiod affects movement behaviour and occu-
pancy dynamics in an amphibian. Freshw Biol 62:1768–1782
Tregenza T, Wedell N (2000) Genetic compatibility, mate choice and
patterns of parentage: invited review. Mol Ecol 9:1013–1027
 Van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004)
Micro-Checker: software for identifying and correcting genotyp-
ing errors in microsatellite data. Mol Ecol Notes 4:535–538
Verkuil YI, Juillet C, Lank DB, Widemo F, Piersma T (2014) Genetic
variation in nuclear and mitochondrial markers supports a large
sex difference in lifetime reproductive skew in a lekking species.
Ecol Evol 4:3626–3632
Wang J (2004) Estimating pairwise relatedness from dominant genetic
markers. Mol Ecol 13(10):3169–3178
http://doc.rero.ch
16
Wang J (2011) COANCESTRY: a program for simulating, estimating
and analysing relatedness and inbreeding coefficients. Mol Ecol
Resour 11:141–145
Wang J (2012) Computationally efficient sibship and parentage assign-
ment from multilocus marker data. Genetics 191(1):183–194
Wegge P, Finne MH, Rolstad J (2007) GPS satellite telemetry provides
new insight into capercaillie Tetrao urogallus brood movements.
Wildl Biol 13:87–94